Title :
TACO - exploiting cluster networks for high-level collective operations
Author :
Nolte, Jörg ; Sato, Mitsuhisa ; Ishikawa, Yutaka
Author_Institution :
Real World Comput. Partnership, Tsukuba, Japan
Abstract :
TACO (Topologies and Collections) is a template library that introduces the flavour of distributed data parallel processing by means of reusable topology classes and C++ templates. The paper introduces TACO´s basic abstractions and provides a performance analysis for basic collective operations on various cluster architectures with several different networks
Keywords :
C++ language; distributed object management; parallel programming; software libraries; software reusability; workstation clusters; C++ templates; TACO; Topologies and Collections; basic abstractions; basic collective operations; cluster architectures; cluster networks; distributed data parallel processing; performance analysis; reusable topology classes; template library; Communication system control; Computer architecture; Computer networks; Concurrent computing; Distributed computing; Ethernet networks; Libraries; Network topology; Performance analysis; Programming profession;
Conference_Titel :
Cluster Computing and the Grid, 2001. Proceedings. First IEEE/ACM International Symposium on
Conference_Location :
Brisbane, Qld.
Print_ISBN :
0-7695-1010-8
DOI :
10.1109/CCGRID.2001.923201
